[Primitive] Floor: Redo the model for floor

[sloorush]

Redo the model function for floor to handle inputs like 1E200.

The current model of floor calculates it by using a very non-intrusive (with the actual primitive being tested) approach by using strings to calculate the result. But it fails to evaluate inputs like 1E200 and fails.

[sloorush]

Turns out ⍕ is a beast of a primitive.

Next steps two approaches:

• The Interfacing approach: try out using .NET interfacing and what impact that has on code coverage and other segments of tests
• The "machine code" approach: implementing floor using binary numbers

[sloorush]

Logging this here for the future, since the PR has been closed

Proposed model in the closed PR was:

    ⍝ model of floor using exponents
    ⍝ has issues with hdbl and hfl values
    modelFloorNew←{
        fmt←{⎕FR≡1287:¯33⍕⍵ ⋄ ¯16⍕⍵}⍵
        (m e)←'E'(≠⊆⊢)fmt
        en←⍎e
        diff←-(⍵<0)∧('.'∊⍕⍵)
        en<0:diff+0
        m↑⍨←3+(⍵<0)+en
        diff+⍎m,'E',e
    }

It does the same thing as the current model but it uses the scientific form the number and removes the decimal places. But it is still not precise.